Step 4. Does Self-Efficacy Mediate the Relationship Between PWES and Workplace Activity?The magnitude of the indirect effect of PWES on PA, through self-efficacy, ranged from .049 to .088 across the three time points. The associated Sobel tests were all significant (p<.002). The change in R2values when PWES was added as the second PWES2PA2SE2PWES2PA2.374(.052).157(.040).315(.051).330(.051)(b)(a)Panel B:6 monthsPWES1PA1SE1PWES1PA1.319(.045).196(.033).447(.044).230(.044)(b)(a)Panel A:BaselinePWES3PA3SE3PWES3PA3.317(.053).139(.041).351(.050).270(.051)(b)(a)Panel C:12 monthsat BRIGHAM YOUNG UNIV on September 13, 2010heb.sagepub.comDownloaded from
176 Health Education & Behavior(April 2010)predictor ranged from .026 to .060. These R2change values were smaller in magnitude than the respective R2values observed when PWES was the only predictor of PA, at all three time points.Although the amount of unique variance shared between PWES and PA was less when self-efficacy was included as a second predictor than it was when PWES predicted PA alone (supporting the role of self-efficacy as a mediator), the direct path between PWES and PA remained statistically significant in all three cross-sectional mediation models. These results suggest that the relationship between PWES and PA is partially mediated by self-efficacy. However, it appears that a substantial portion of the relationship between PWES and PA is not explained by the indirect effect through self-efficacy alone.Longitudinal AnalysesBecause temporal sequencing can strengthen the implication of causative mediation processes (Bauman et al., 2002), we included a mediation model to explore the potential longitudinal relationships among PWES, self-efficacy, and PA. Figure 2 shows our longitudinal model, whereby PWES at baseline was hypothesized to predict PA at 12 months, with self-efficacy at 6 months hypothesized to serve as a significant mediator. Structural path coefficients and associated standard errors are indicated in the figure.Table 1 includes the magnitude of the mediating effect and the associated result of the Sobel test of significance, the overall shared variance between PWES at baseline and PA at 12 months (i.e., R2when PA at 12 months is regressed on baseline PWES alone), and the unique shared variance between baseline PWES and PA at 12 months in the mediation model (i.e., the change in R2when baseline PWES is added as a second predictor to a model with PA at 12 months already regressed on self-efficacyat6 months).Table 1. Indirect SE Mediation Effect Size, Sobel zTest Statistic, Unmediated Direct PWES Effect Size, and Attenuated Direct PWES Effect Size for Cross-Sectional, Longitudinal, and Residual Change Mediation ModelMediation Path (a)( b) Path (c) Model Effect Size Sobel z Path (c) R2 R2ChangeCross-sectional analysesBaseline .088 5.13 (p<.001) .053 .0266 months .049 3.31 (p<.001) .079 .06012 months .049 3.05 (p=.002) .057 .040Longitudinal analysisLongitudinal .039 2.66 (p=.008) .029 .022Residual change analysesInitial change .021 2.05 (p=.040) .031 .027Delayed change .012 1.18 (p=.237) .018 .015Overall change .022 2.02 (p=.044) .028 .024NOTE: SE =self-efficacy; PWES =perceptions of workplace environment.